xref: /dragonfly/sys/netinet6/ip6_input.c (revision cb3deea5)
1 /*	$FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.15 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/kernel.h>
80 #include <sys/syslog.h>
81 #include <sys/proc.h>
82 #include <sys/priv.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_dl.h>
87 #include <net/route.h>
88 #include <net/netisr.h>
89 #include <net/pfil.h>
90 
91 #include <sys/thread2.h>
92 #include <sys/msgport2.h>
93 #include <net/netmsg2.h>
94 #include <net/netisr2.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_systm.h>
98 #ifdef INET
99 #include <netinet/ip.h>
100 #include <netinet/ip_icmp.h>
101 #endif /* INET */
102 #include <netinet/ip6.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 
111 #ifdef IPSEC
112 #include <netinet6/ipsec.h>
113 #ifdef INET6
114 #include <netinet6/ipsec6.h>
115 #endif
116 #endif
117 
118 #ifdef FAST_IPSEC
119 #include <netproto/ipsec/ipsec.h>
120 #include <netproto/ipsec/ipsec6.h>
121 #define	IPSEC
122 #endif /* FAST_IPSEC */
123 
124 #include <net/ip6fw/ip6_fw.h>
125 
126 #include <netinet6/ip6protosw.h>
127 
128 #include <net/net_osdep.h>
129 
130 extern struct domain inet6domain;
131 extern struct protosw inet6sw[];
132 
133 u_char ip6_protox[IPPROTO_MAX];
134 struct in6_ifaddr *in6_ifaddr;
135 
136 int ip6_forward_srcrt;			/* XXX */
137 int ip6_sourcecheck;			/* XXX */
138 int ip6_sourcecheck_interval;		/* XXX */
139 
140 int ip6_ours_check_algorithm;
141 
142 struct pfil_head inet6_pfil_hook;
143 
144 /* firewall hooks */
145 ip6_fw_chk_t *ip6_fw_chk_ptr;
146 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
147 int ip6_fw_enable = 1;
148 
149 struct ip6stat ip6stat;
150 
151 static void ip6_init2 (void *);
152 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
153 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
154 static void ip6_input(netmsg_t msg);
155 #ifdef PULLDOWN_TEST
156 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
157 #endif
158 static void transport6_processing_handler(netmsg_t netmsg);
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
165 ip6_init(void)
166 {
167 	struct protosw *pr;
168 	int i;
169 	struct timeval tv;
170 
171 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
172 	if (pr == NULL)
173 		panic("ip6_init");
174 	for (i = 0; i < IPPROTO_MAX; i++)
175 		ip6_protox[i] = pr - inet6sw;
176 	for (pr = inet6domain.dom_protosw;
177 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
178 		if (pr->pr_domain->dom_family == PF_INET6 &&
179 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
180 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
181 
182 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
183 	inet6_pfil_hook.ph_af = AF_INET6;
184 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
185 		kprintf("%s: WARNING: unable to register pfil hook, "
186 			"error %d\n", __func__, i);
187 	}
188 
189 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX hashfn */
190 	scope6_init();
191 	addrsel_policy_init();
192 	nd6_init();
193 	frag6_init();
194 	/*
195 	 * in many cases, random() here does NOT return random number
196 	 * as initialization during bootstrap time occur in fixed order.
197 	 */
198 	microtime(&tv);
199 	ip6_flow_seq = krandom() ^ tv.tv_usec;
200 	microtime(&tv);
201 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
202 }
203 
204 static void
205 ip6_init2(void *dummy)
206 {
207 	nd6_timer_init();
208 	in6_tmpaddrtimer_init();
209 }
210 
211 /* cheat */
212 /* This must be after route_init(), which is now SI_ORDER_THIRD */
213 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
214 
215 extern struct	route_in6 ip6_forward_rt;
216 
217 static void
218 ip6_input(netmsg_t msg)
219 {
220 	struct mbuf *m = msg->packet.nm_packet;
221 	struct ip6_hdr *ip6;
222 	int off = sizeof(struct ip6_hdr), nest;
223 	u_int32_t plen;
224 	u_int32_t rtalert = ~0;
225 	int nxt, ours = 0, rh_present = 0;
226 	struct ifnet *deliverifp = NULL;
227 	struct in6_addr odst;
228 	int srcrt = 0;
229 
230 #ifdef IPSEC
231 	/*
232 	 * should the inner packet be considered authentic?
233 	 * see comment in ah4_input().
234 	 */
235 	if (m) {
236 		m->m_flags &= ~M_AUTHIPHDR;
237 		m->m_flags &= ~M_AUTHIPDGM;
238 	}
239 #endif
240 
241 	/*
242 	 * make sure we don't have onion peering information into m_aux.
243 	 */
244 	ip6_delaux(m);
245 
246 	/*
247 	 * mbuf statistics
248 	 */
249 	if (m->m_flags & M_EXT) {
250 		if (m->m_next)
251 			ip6stat.ip6s_mext2m++;
252 		else
253 			ip6stat.ip6s_mext1++;
254 	} else {
255 #define M2MMAX	NELEM(ip6stat.ip6s_m2m)
256 		if (m->m_next) {
257 			if (m->m_flags & M_LOOP) {
258 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
259 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
260 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
261 			else
262 				ip6stat.ip6s_m2m[0]++;
263 		} else
264 			ip6stat.ip6s_m1++;
265 #undef M2MMAX
266 	}
267 
268 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
269 	ip6stat.ip6s_total++;
270 
271 #ifndef PULLDOWN_TEST
272 	/*
273 	 * L2 bridge code and some other code can return mbuf chain
274 	 * that does not conform to KAME requirement.  too bad.
275 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
276 	 */
277 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
278 		struct mbuf *n;
279 
280 		n = m_getb(m->m_pkthdr.len, M_NOWAIT, MT_HEADER, M_PKTHDR);
281 		if (n == NULL)
282 			goto bad;
283 		M_MOVE_PKTHDR(n, m);
284 
285 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
286 		n->m_len = n->m_pkthdr.len;
287 		m_freem(m);
288 		m = n;
289 	}
290 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
291 #endif
292 
293 	if (m->m_len < sizeof(struct ip6_hdr)) {
294 		struct ifnet *inifp;
295 		inifp = m->m_pkthdr.rcvif;
296 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
297 			ip6stat.ip6s_toosmall++;
298 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
299 			goto bad2;
300 		}
301 	}
302 
303 	ip6 = mtod(m, struct ip6_hdr *);
304 
305 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
306 		ip6stat.ip6s_badvers++;
307 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
308 		goto bad;
309 	}
310 
311 	/*
312 	 * Run through list of hooks for input packets.
313 	 *
314 	 * NB: Beware of the destination address changing
315 	 *     (e.g. by NAT rewriting). When this happens,
316 	 *     tell ip6_forward to do the right thing.
317 	 */
318 	if (pfil_has_hooks(&inet6_pfil_hook)) {
319 		odst = ip6->ip6_dst;
320 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
321 		    m->m_pkthdr.rcvif, PFIL_IN)) {
322 			goto bad2;
323 		}
324 		if (m == NULL)			/* consumed by filter */
325 			goto bad2;
326 		ip6 = mtod(m, struct ip6_hdr *);
327 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
328 	}
329 
330 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
331 
332 #ifdef ALTQ
333 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
334 		/* packet is dropped by traffic conditioner */
335 		return;
336 	}
337 #endif
338 
339 	/*
340 	 * Check with the firewall...
341 	 */
342 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
343 		u_short port = 0;
344 		/* If ipfw says divert, we have to just drop packet */
345 		/* use port as a dummy argument */
346 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
347 			m_freem(m);
348 			m = NULL;
349 		}
350 		if (!m)
351 			goto bad2;
352 	}
353 
354 	/*
355 	 * Check against address spoofing/corruption.
356 	 */
357 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
358 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
359 		/*
360 		 * XXX: "badscope" is not very suitable for a multicast source.
361 		 */
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
367 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
368 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
369 		ip6stat.ip6s_badscope++;
370 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
371 		goto bad;
372 	}
373 
374 	/*
375 	 * The following check is not documented in specs.  A malicious
376 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
377 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
378 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
379 	 *
380 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
381 	 * support IPv4-less kernel compilation, we cannot support SIIT
382 	 * environment at all.  So, it makes more sense for us to reject any
383 	 * malicious packets for non-SIIT environment, than try to do a
384 	 * partial support for SIIT environment.
385 	 */
386 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
387 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
388 		ip6stat.ip6s_badscope++;
389 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
390 		goto bad;
391 	}
392 #if 0
393 	/*
394 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
395 	 *
396 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
397 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
398 	 * is revised to forbid relaying case.
399 	 */
400 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
401 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
402 		ip6stat.ip6s_badscope++;
403 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
404 		goto bad;
405 	}
406 #endif
407 
408 	/* drop packets if interface ID portion is already filled */
409 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
410 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
411 		    ip6->ip6_src.s6_addr16[1]) {
412 			ip6stat.ip6s_badscope++;
413 			goto bad;
414 		}
415 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
416 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
417 		    ip6->ip6_dst.s6_addr16[1]) {
418 			ip6stat.ip6s_badscope++;
419 			goto bad;
420 		}
421 	}
422 
423 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
424 		ip6->ip6_src.s6_addr16[1]
425 			= htons(m->m_pkthdr.rcvif->if_index);
426 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
427 		ip6->ip6_dst.s6_addr16[1]
428 			= htons(m->m_pkthdr.rcvif->if_index);
429 
430 	/*
431 	 * Multicast check
432 	 *
433 	 * WARNING!  For general subnet proxying the interface hw will
434 	 *	     likely filter out the multicast solicitations, so
435 	 *	     the interface must be in promiscuous mode.
436 	 */
437 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
438 		ours = 1;
439 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
440 		deliverifp = m->m_pkthdr.rcvif;
441 		goto hbhcheck;
442 	}
443 
444 	/*
445 	 *  Unicast check
446 	 */
447 	if (ip6_forward_rt.ro_rt != NULL &&
448 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
449 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
450 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr)) {
451 		ip6stat.ip6s_forward_cachehit++;
452 	} else {
453 		struct sockaddr_in6 *dst6;
454 
455 		if (ip6_forward_rt.ro_rt) {
456 			/* route is down or destination is different */
457 			ip6stat.ip6s_forward_cachemiss++;
458 			RTFREE(ip6_forward_rt.ro_rt);
459 			ip6_forward_rt.ro_rt = 0;
460 		}
461 
462 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
463 		dst6 = &ip6_forward_rt.ro_dst;
464 		dst6->sin6_len = sizeof(struct sockaddr_in6);
465 		dst6->sin6_family = AF_INET6;
466 		dst6->sin6_addr = ip6->ip6_dst;
467 
468 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
469 	}
470 
471 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
472 
473 	/*
474 	 * Accept the packet if the forwarding interface to the destination
475 	 * according to the routing table is the loopback interface,
476 	 * unless the associated route has a gateway.
477 	 * Note that this approach causes to accept a packet if there is a
478 	 * route to the loopback interface for the destination of the packet.
479 	 * But we think it's even useful in some situations, e.g. when using
480 	 * a special daemon which wants to intercept the packet.
481 	 *
482 	 * XXX: some OSes automatically make a cloned route for the destination
483 	 * of an outgoing packet.  If the outgoing interface of the packet
484 	 * is a loopback one, the kernel would consider the packet to be
485 	 * accepted, even if we have no such address assinged on the interface.
486 	 * We check the cloned flag of the route entry to reject such cases,
487 	 * assuming that route entries for our own addresses are not made by
488 	 * cloning (it should be true because in6_addloop explicitly installs
489 	 * the host route).  However, we might have to do an explicit check
490 	 * while it would be less efficient.  Or, should we rather install a
491 	 * reject route for such a case?
492 	 */
493 	if (ip6_forward_rt.ro_rt &&
494 	    (ip6_forward_rt.ro_rt->rt_flags &
495 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
496 #ifdef RTF_WASCLONED
497 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
498 #endif
499 #ifdef RTF_CLONED
500 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
501 #endif
502 #if 0
503 	    /*
504 	     * The check below is redundant since the comparison of
505 	     * the destination and the key of the rtentry has
506 	     * already done through looking up the routing table.
507 	     */
508 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
509 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
510 #endif
511 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
512 		struct in6_ifaddr *ia6 =
513 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
514 
515 		/*
516 		 * record address information into m_aux.
517 		 */
518 		ip6_setdstifaddr(m, ia6);
519 
520 		/*
521 		 * packets to a tentative, duplicated, or somehow invalid
522 		 * address must not be accepted.
523 		 */
524 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
525 			/* this address is ready */
526 			ours = 1;
527 			deliverifp = ia6->ia_ifp;	/* correct? */
528 			/* Count the packet in the ip address stats */
529 			IFA_STAT_INC(&ia6->ia_ifa, ipackets, 1);
530 			IFA_STAT_INC(&ia6->ia_ifa, ibytes, m->m_pkthdr.len);
531 			goto hbhcheck;
532 		} else {
533 			/* address is not ready, so discard the packet. */
534 			nd6log((LOG_INFO,
535 			    "ip6_input: packet to an unready address %s->%s\n",
536 			    ip6_sprintf(&ip6->ip6_src),
537 			    ip6_sprintf(&ip6->ip6_dst)));
538 
539 			goto bad;
540 		}
541 	}
542 
543 	/*
544 	 * FAITH (Firewall Aided Internet Translator)
545 	 */
546 	if (ip6_keepfaith) {
547 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
548 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
549 			/* XXX do we need more sanity checks? */
550 			ours = 1;
551 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
552 			goto hbhcheck;
553 		}
554 	}
555 
556 	/*
557 	 * Now there is no reason to process the packet if it's not our own
558 	 * and we're not a router.
559 	 */
560 	if (!ip6_forwarding) {
561 		ip6stat.ip6s_cantforward++;
562 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
563 		goto bad;
564 	}
565 
566 hbhcheck:
567 	/*
568 	 * record address information into m_aux, if we don't have one yet.
569 	 * note that we are unable to record it, if the address is not listed
570 	 * as our interface address (e.g. multicast addresses, addresses
571 	 * within FAITH prefixes and such).
572 	 */
573 	if (deliverifp && !ip6_getdstifaddr(m)) {
574 		struct in6_ifaddr *ia6;
575 
576 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
577 		if (ia6) {
578 			if (!ip6_setdstifaddr(m, ia6)) {
579 				/*
580 				 * XXX maybe we should drop the packet here,
581 				 * as we could not provide enough information
582 				 * to the upper layers.
583 				 */
584 			}
585 		}
586 	}
587 
588 	/*
589 	 * Process Hop-by-Hop options header if it's contained.
590 	 * m may be modified in ip6_hopopts_input().
591 	 * If a JumboPayload option is included, plen will also be modified.
592 	 */
593 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
594 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
595 		struct ip6_hbh *hbh;
596 
597 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
598 #if 0	/*touches NULL pointer*/
599 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
600 #endif
601 			goto bad2;	/* m have already been freed */
602 		}
603 
604 		/* adjust pointer */
605 		ip6 = mtod(m, struct ip6_hdr *);
606 
607 		/*
608 		 * if the payload length field is 0 and the next header field
609 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
610 		 * option MUST be included.
611 		 */
612 		if (ip6->ip6_plen == 0 && plen == 0) {
613 			/*
614 			 * Note that if a valid jumbo payload option is
615 			 * contained, ip6_hoptops_input() must set a valid
616 			 * (non-zero) payload length to the variable plen.
617 			 */
618 			ip6stat.ip6s_badoptions++;
619 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
620 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
621 			icmp6_error(m, ICMP6_PARAM_PROB,
622 				    ICMP6_PARAMPROB_HEADER,
623 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
624 			goto bad2;
625 		}
626 #ifndef PULLDOWN_TEST
627 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
628 		hbh = (struct ip6_hbh *)(ip6 + 1);
629 #else
630 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
631 			sizeof(struct ip6_hbh));
632 		if (hbh == NULL) {
633 			ip6stat.ip6s_tooshort++;
634 			goto bad2;
635 		}
636 #endif
637 		nxt = hbh->ip6h_nxt;
638 
639 		/*
640 		 * If we are acting as a router and the packet contains a
641 		 * router alert option, see if we know the option value.
642 		 * Currently, we only support the option value for MLD, in which
643 		 * case we should pass the packet to the multicast routing
644 		 * daemon.
645 		 */
646 		if (rtalert != ~0 && ip6_forwarding) {
647 			switch (rtalert) {
648 			case IP6OPT_RTALERT_MLD:
649 				ours = 1;
650 				break;
651 			default:
652 				/*
653 				 * RFC2711 requires unrecognized values must be
654 				 * silently ignored.
655 				 */
656 				break;
657 			}
658 		}
659 	} else
660 		nxt = ip6->ip6_nxt;
661 
662 	/*
663 	 * Check that the amount of data in the buffers
664 	 * is as at least much as the IPv6 header would have us expect.
665 	 * Trim mbufs if longer than we expect.
666 	 * Drop packet if shorter than we expect.
667 	 */
668 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
669 		ip6stat.ip6s_tooshort++;
670 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
671 		goto bad;
672 	}
673 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
674 		if (m->m_len == m->m_pkthdr.len) {
675 			m->m_len = sizeof(struct ip6_hdr) + plen;
676 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
677 		} else
678 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
679 	}
680 
681 	/*
682 	 * Forward if desirable.
683 	 */
684 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
685 		/*
686 		 * If we are acting as a multicast router, all
687 		 * incoming multicast packets are passed to the
688 		 * kernel-level multicast forwarding function.
689 		 * The packet is returned (relatively) intact; if
690 		 * ip6_mforward() returns a non-zero value, the packet
691 		 * must be discarded, else it may be accepted below.
692 		 */
693 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
694 			ip6stat.ip6s_cantforward++;
695 			goto bad;
696 		}
697 		if (!ours)
698 			goto bad;
699 	} else if (!ours) {
700 		ip6_forward(m, srcrt);
701 		goto bad2;
702 	}
703 
704 	ip6 = mtod(m, struct ip6_hdr *);
705 
706 	/*
707 	 * Malicious party may be able to use IPv4 mapped addr to confuse
708 	 * tcp/udp stack and bypass security checks (act as if it was from
709 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
710 	 *
711 	 * For SIIT end node behavior, you may want to disable the check.
712 	 * However, you will  become vulnerable to attacks using IPv4 mapped
713 	 * source.
714 	 */
715 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
716 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
717 		ip6stat.ip6s_badscope++;
718 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
719 		goto bad;
720 	}
721 
722 	/*
723 	 * Tell launch routine the next header
724 	 */
725 	ip6stat.ip6s_delivered++;
726 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
727 	nest = 0;
728 
729 	rh_present = 0;
730 	while (nxt != IPPROTO_DONE) {
731 		struct protosw *sw6;
732 
733 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
734 			ip6stat.ip6s_toomanyhdr++;
735 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
736 			goto bad;
737 		}
738 
739 		/*
740 		 * protection against faulty packet - there should be
741 		 * more sanity checks in header chain processing.
742 		 */
743 		if (m->m_pkthdr.len < off) {
744 			ip6stat.ip6s_tooshort++;
745 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
746 			goto bad;
747 		}
748 
749 #if 0
750 		/*
751 		 * do we need to do it for every header?  yeah, other
752 		 * functions can play with it (like re-allocate and copy).
753 		 */
754 		mhist = ip6_addaux(m);
755 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
756 			hist = mtod(mhist, caddr_t) + mhist->m_len;
757 			bcopy(&nxt, hist, sizeof(nxt));
758 			mhist->m_len += sizeof(nxt);
759 		} else {
760 			ip6stat.ip6s_toomanyhdr++;
761 			goto bad;
762 		}
763 #endif
764 
765 		if (nxt == IPPROTO_ROUTING) {
766 			if (rh_present++) {
767 				in6_ifstat_inc(m->m_pkthdr.rcvif,
768 				    ifs6_in_hdrerr);
769 				ip6stat.ip6s_badoptions++;
770 				goto bad;
771 			}
772 		}
773 
774 		sw6 = &inet6sw[ip6_protox[nxt]];
775 #ifdef IPSEC
776 		/*
777 		 * enforce IPsec policy checking if we are seeing last header.
778 		 * note that we do not visit this with protocols with pcb layer
779 		 * code - like udp/tcp/raw ip.
780 		 */
781 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
782 			ipsec6stat.in_polvio++;
783 			goto bad;
784 		}
785 #endif
786 		/*
787 		 * If this is a terminal header forward to the port, otherwise
788 		 * process synchronously for more headers.
789 		 */
790 		if (sw6->pr_flags & PR_LASTHDR) {
791 			struct netmsg_packet *pmsg;
792 			lwkt_port_t port;
793 
794 			port = netisr_cpuport(0); /* XXX */
795 			KKASSERT(port != NULL);
796 			pmsg = &m->m_hdr.mh_netmsg;
797 			netmsg_init(&pmsg->base, NULL,
798 				    &netisr_apanic_rport,
799 				    0, transport6_processing_handler);
800 			pmsg->nm_packet = m;
801 			pmsg->nm_nxt = nxt;
802 			pmsg->base.lmsg.u.ms_result = off;
803 			lwkt_sendmsg(port, &pmsg->base.lmsg);
804 			/* done with m */
805 			nxt = IPPROTO_DONE;
806 		} else {
807 			nxt = sw6->pr_input(&m, &off, nxt);
808 		}
809 	}
810 	goto bad2;
811 bad:
812 	m_freem(m);
813 bad2:
814 	;
815 	/* msg was embedded in the mbuf, do not reply! */
816 }
817 
818 /*
819  * We have to call the pr_input() function from the correct protocol
820  * thread.  The sw6->pr_soport() request at the end of ip6_input()
821  * returns the port and we forward a netmsg to the port to execute
822  * this function.
823  */
824 static void
825 transport6_processing_handler(netmsg_t netmsg)
826 {
827 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
828 	struct protosw *sw6;
829 	int hlen;
830 	int nxt;
831 
832 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
833 	hlen = pmsg->base.lmsg.u.ms_result;
834 
835 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
836 	KKASSERT(nxt == IPPROTO_DONE);
837 	/* netmsg was embedded in the mbuf, do not reply! */
838 }
839 
840 /*
841  * set/grab in6_ifaddr correspond to IPv6 destination address.
842  * XXX backward compatibility wrapper
843  */
844 static struct ip6aux *
845 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
846 {
847 	struct ip6aux *n;
848 
849 	n = ip6_addaux(m);
850 	if (n)
851 		n->ip6a_dstia6 = ia6;
852 	return n;	/* NULL if failed to set */
853 }
854 
855 struct in6_ifaddr *
856 ip6_getdstifaddr(struct mbuf *m)
857 {
858 	struct ip6aux *n;
859 
860 	n = ip6_findaux(m);
861 	if (n)
862 		return n->ip6a_dstia6;
863 	else
864 		return NULL;
865 }
866 
867 /*
868  * Hop-by-Hop options header processing. If a valid jumbo payload option is
869  * included, the real payload length will be stored in plenp.
870  */
871 static int
872 ip6_hopopts_input(u_int32_t *plenp,
873 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
874 		  struct mbuf **mp,
875 		  int *offp)
876 {
877 	struct mbuf *m = *mp;
878 	int off = *offp, hbhlen;
879 	struct ip6_hbh *hbh;
880 
881 	/* validation of the length of the header */
882 #ifndef PULLDOWN_TEST
883 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
884 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
885 	hbhlen = (hbh->ip6h_len + 1) << 3;
886 
887 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
888 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
889 #else
890 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
891 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
892 	if (hbh == NULL) {
893 		ip6stat.ip6s_tooshort++;
894 		return -1;
895 	}
896 	hbhlen = (hbh->ip6h_len + 1) << 3;
897 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
898 		hbhlen);
899 	if (hbh == NULL) {
900 		ip6stat.ip6s_tooshort++;
901 		return -1;
902 	}
903 #endif
904 	off += hbhlen;
905 	hbhlen -= sizeof(struct ip6_hbh);
906 
907 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
908 				hbhlen, rtalertp, plenp) < 0)
909 		return (-1);
910 
911 	*offp = off;
912 	*mp = m;
913 	return (0);
914 }
915 
916 /*
917  * Search header for all Hop-by-hop options and process each option.
918  * This function is separate from ip6_hopopts_input() in order to
919  * handle a case where the sending node itself process its hop-by-hop
920  * options header. In such a case, the function is called from ip6_output().
921  *
922  * The function assumes that hbh header is located right after the IPv6 header
923  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
924  * opthead + hbhlen is located in continuous memory region.
925  */
926 int
927 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
928 		    u_int32_t *rtalertp, u_int32_t *plenp)
929 {
930 	struct ip6_hdr *ip6;
931 	int optlen = 0;
932 	u_int8_t *opt = opthead;
933 	u_int16_t rtalert_val;
934 	u_int32_t jumboplen;
935 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
936 
937 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
938 		switch (*opt) {
939 		case IP6OPT_PAD1:
940 			optlen = 1;
941 			break;
942 		case IP6OPT_PADN:
943 			if (hbhlen < IP6OPT_MINLEN) {
944 				ip6stat.ip6s_toosmall++;
945 				goto bad;
946 			}
947 			optlen = *(opt + 1) + 2;
948 			break;
949 		case IP6OPT_RTALERT:
950 			/* XXX may need check for alignment */
951 			if (hbhlen < IP6OPT_RTALERT_LEN) {
952 				ip6stat.ip6s_toosmall++;
953 				goto bad;
954 			}
955 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
956 				/* XXX stat */
957 				icmp6_error(m, ICMP6_PARAM_PROB,
958 				    ICMP6_PARAMPROB_HEADER,
959 				    erroff + opt + 1 - opthead);
960 				return (-1);
961 			}
962 			optlen = IP6OPT_RTALERT_LEN;
963 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
964 			*rtalertp = ntohs(rtalert_val);
965 			break;
966 		case IP6OPT_JUMBO:
967 			/* XXX may need check for alignment */
968 			if (hbhlen < IP6OPT_JUMBO_LEN) {
969 				ip6stat.ip6s_toosmall++;
970 				goto bad;
971 			}
972 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
973 				/* XXX stat */
974 				icmp6_error(m, ICMP6_PARAM_PROB,
975 				    ICMP6_PARAMPROB_HEADER,
976 				    erroff + opt + 1 - opthead);
977 				return (-1);
978 			}
979 			optlen = IP6OPT_JUMBO_LEN;
980 
981 			/*
982 			 * IPv6 packets that have non 0 payload length
983 			 * must not contain a jumbo payload option.
984 			 */
985 			ip6 = mtod(m, struct ip6_hdr *);
986 			if (ip6->ip6_plen) {
987 				ip6stat.ip6s_badoptions++;
988 				icmp6_error(m, ICMP6_PARAM_PROB,
989 				    ICMP6_PARAMPROB_HEADER,
990 				    erroff + opt - opthead);
991 				return (-1);
992 			}
993 
994 			/*
995 			 * We may see jumbolen in unaligned location, so
996 			 * we'd need to perform bcopy().
997 			 */
998 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
999 			jumboplen = (u_int32_t)htonl(jumboplen);
1000 
1001 #if 1
1002 			/*
1003 			 * if there are multiple jumbo payload options,
1004 			 * *plenp will be non-zero and the packet will be
1005 			 * rejected.
1006 			 * the behavior may need some debate in ipngwg -
1007 			 * multiple options does not make sense, however,
1008 			 * there's no explicit mention in specification.
1009 			 */
1010 			if (*plenp != 0) {
1011 				ip6stat.ip6s_badoptions++;
1012 				icmp6_error(m, ICMP6_PARAM_PROB,
1013 				    ICMP6_PARAMPROB_HEADER,
1014 				    erroff + opt + 2 - opthead);
1015 				return (-1);
1016 			}
1017 #endif
1018 
1019 			/*
1020 			 * jumbo payload length must be larger than 65535.
1021 			 */
1022 			if (jumboplen <= IPV6_MAXPACKET) {
1023 				ip6stat.ip6s_badoptions++;
1024 				icmp6_error(m, ICMP6_PARAM_PROB,
1025 				    ICMP6_PARAMPROB_HEADER,
1026 				    erroff + opt + 2 - opthead);
1027 				return (-1);
1028 			}
1029 			*plenp = jumboplen;
1030 
1031 			break;
1032 		default:		/* unknown option */
1033 			if (hbhlen < IP6OPT_MINLEN) {
1034 				ip6stat.ip6s_toosmall++;
1035 				goto bad;
1036 			}
1037 			optlen = ip6_unknown_opt(opt, m,
1038 			    erroff + opt - opthead);
1039 			if (optlen == -1)
1040 				return (-1);
1041 			optlen += 2;
1042 			break;
1043 		}
1044 	}
1045 
1046 	return (0);
1047 
1048 bad:
1049 	m_freem(m);
1050 	return (-1);
1051 }
1052 
1053 /*
1054  * Unknown option processing.
1055  * The third argument `off' is the offset from the IPv6 header to the option,
1056  * which is necessary if the IPv6 header the and option header and IPv6 header
1057  * is not continuous in order to return an ICMPv6 error.
1058  */
1059 int
1060 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1061 {
1062 	struct ip6_hdr *ip6;
1063 
1064 	switch (IP6OPT_TYPE(*optp)) {
1065 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1066 		return ((int)*(optp + 1));
1067 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1068 		m_freem(m);
1069 		return (-1);
1070 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1071 		ip6stat.ip6s_badoptions++;
1072 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1073 		return (-1);
1074 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1075 		ip6stat.ip6s_badoptions++;
1076 		ip6 = mtod(m, struct ip6_hdr *);
1077 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1078 		    (m->m_flags & (M_BCAST|M_MCAST)))
1079 			m_freem(m);
1080 		else
1081 			icmp6_error(m, ICMP6_PARAM_PROB,
1082 				    ICMP6_PARAMPROB_OPTION, off);
1083 		return (-1);
1084 	}
1085 
1086 	m_freem(m);		/* XXX: NOTREACHED */
1087 	return (-1);
1088 }
1089 
1090 /*
1091  * Create the "control" list for this pcb.
1092  * The function will not modify mbuf chain at all.
1093  *
1094  * with KAME mbuf chain restriction:
1095  * The routine will be called from upper layer handlers like tcp6_input().
1096  * Thus the routine assumes that the caller (tcp6_input) have already
1097  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1098  * very first mbuf on the mbuf chain.
1099  */
1100 void
1101 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1102 		struct mbuf *m)
1103 {
1104 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1105 	struct thread *td = curthread;	/* XXX */
1106 	int privileged = 0;
1107 
1108 
1109 	if (priv_check(td, PRIV_ROOT) == 0)
1110 		privileged++;
1111 
1112 #ifdef SO_TIMESTAMP
1113 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1114 		struct timeval tv;
1115 
1116 		microtime(&tv);
1117 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1118 		    SCM_TIMESTAMP, SOL_SOCKET);
1119 		if (*mp) {
1120 			mp = &(*mp)->m_next;
1121 		}
1122 	}
1123 #endif
1124 
1125 	/* RFC 2292 sec. 5 */
1126 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1127 		struct in6_pktinfo pi6;
1128 
1129 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1130 		in6_clearscope(&pi6.ipi6_addr); /* XXX */
1131 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif) ?
1132 		    m->m_pkthdr.rcvif->if_index : 0;
1133 		*mp = sbcreatecontrol((caddr_t)&pi6, sizeof(struct in6_pktinfo),
1134 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1135 		if (*mp)
1136 			mp = &(*mp)->m_next;
1137 	}
1138 
1139 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1140 		int hlim = ip6->ip6_hlim & 0xff;
1141 
1142 		*mp = sbcreatecontrol((caddr_t)&hlim, sizeof(int),
1143 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1144 		if (*mp)
1145 			mp = &(*mp)->m_next;
1146 	}
1147 
1148 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1149 		u_int32_t flowinfo;
1150 		int tclass;
1151 
1152 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1153 		flowinfo >>= 20;
1154 
1155 		tclass = flowinfo & 0xff;
1156 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1157 		    IPV6_TCLASS, IPPROTO_IPV6);
1158 		if (*mp)
1159 			mp = &(*mp)->m_next;
1160 	}
1161 
1162 	/*
1163 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1164 	 * for the option, but it might be too strict, since there might
1165 	 * be some hop-by-hop options which can be returned to normal user.
1166 	 * See RFC 2292 section 6.
1167 	 */
1168 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1169 		/*
1170 		 * Check if a hop-by-hop options header is contatined in the
1171 		 * received packet, and if so, store the options as ancillary
1172 		 * data. Note that a hop-by-hop options header must be
1173 		 * just after the IPv6 header, which is assured through the
1174 		 * IPv6 input processing.
1175 		 */
1176 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1177 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1178 			struct ip6_hbh *hbh;
1179 			int hbhlen = 0;
1180 #ifdef PULLDOWN_TEST
1181 			struct mbuf *ext;
1182 #endif
1183 
1184 #ifndef PULLDOWN_TEST
1185 			hbh = (struct ip6_hbh *)(ip6 + 1);
1186 			hbhlen = (hbh->ip6h_len + 1) << 3;
1187 #else
1188 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1189 			    ip6->ip6_nxt);
1190 			if (ext == NULL) {
1191 				ip6stat.ip6s_tooshort++;
1192 				return;
1193 			}
1194 			hbh = mtod(ext, struct ip6_hbh *);
1195 			hbhlen = (hbh->ip6h_len + 1) << 3;
1196 			if (hbhlen != ext->m_len) {
1197 				m_freem(ext);
1198 				ip6stat.ip6s_tooshort++;
1199 				return;
1200 			}
1201 #endif
1202 
1203 			/*
1204 			 * XXX: We copy the whole header even if a
1205 			 * jumbo payload option is included, the option which
1206 			 * is to be removed before returning according to
1207 			 * RFC 2292.
1208 			 * Note: this constraint is removed in 2292bis.
1209 			 */
1210 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1211 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1212 			    IPPROTO_IPV6);
1213 			if (*mp)
1214 				mp = &(*mp)->m_next;
1215 #ifdef PULLDOWN_TEST
1216 			m_freem(ext);
1217 #endif
1218 		}
1219 	}
1220 
1221 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1222 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1223 		int proto, off, nxt;
1224 
1225 		/*
1226 		 * go through the header chain to see if a routing header is
1227 		 * contained in the packet. We need this information to store
1228 		 * destination options headers (if any) properly.
1229 		 * XXX: performance issue. We should record this info when
1230 		 * processing extension headers in incoming routine.
1231 		 * (todo) use m_aux?
1232 		 */
1233 		proto = IPPROTO_IPV6;
1234 		off = 0;
1235 		nxt = -1;
1236 		while (1) {
1237 			int newoff;
1238 
1239 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1240 			if (newoff < 0)
1241 				break;
1242 			if (newoff < off) /* invalid, check for safety */
1243 				break;
1244 			if ((proto = nxt) == IPPROTO_ROUTING)
1245 				break;
1246 			off = newoff;
1247 		}
1248 	}
1249 
1250 	if ((in6p->in6p_flags &
1251 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1252 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1253 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1254 
1255 		/*
1256 		 * Search for destination options headers or routing
1257 		 * header(s) through the header chain, and stores each
1258 		 * header as ancillary data.
1259 		 * Note that the order of the headers remains in
1260 		 * the chain of ancillary data.
1261 		 */
1262 		while (1) {	/* is explicit loop prevention necessary? */
1263 			struct ip6_ext *ip6e = NULL;
1264 			int elen;
1265 #ifdef PULLDOWN_TEST
1266 			struct mbuf *ext = NULL;
1267 #endif
1268 
1269 			/*
1270 			 * if it is not an extension header, don't try to
1271 			 * pull it from the chain.
1272 			 */
1273 			switch (nxt) {
1274 			case IPPROTO_DSTOPTS:
1275 			case IPPROTO_ROUTING:
1276 			case IPPROTO_HOPOPTS:
1277 			case IPPROTO_AH: /* is it possible? */
1278 				break;
1279 			default:
1280 				goto loopend;
1281 			}
1282 
1283 #ifndef PULLDOWN_TEST
1284 			if (off + sizeof(*ip6e) > m->m_len)
1285 				goto loopend;
1286 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1287 			if (nxt == IPPROTO_AH)
1288 				elen = (ip6e->ip6e_len + 2) << 2;
1289 			else
1290 				elen = (ip6e->ip6e_len + 1) << 3;
1291 			if (off + elen > m->m_len)
1292 				goto loopend;
1293 #else
1294 			ext = ip6_pullexthdr(m, off, nxt);
1295 			if (ext == NULL) {
1296 				ip6stat.ip6s_tooshort++;
1297 				return;
1298 			}
1299 			ip6e = mtod(ext, struct ip6_ext *);
1300 			if (nxt == IPPROTO_AH)
1301 				elen = (ip6e->ip6e_len + 2) << 2;
1302 			else
1303 				elen = (ip6e->ip6e_len + 1) << 3;
1304 			if (elen != ext->m_len) {
1305 				m_freem(ext);
1306 				ip6stat.ip6s_tooshort++;
1307 				return;
1308 			}
1309 #endif
1310 
1311 			switch (nxt) {
1312 			case IPPROTO_DSTOPTS:
1313 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1314 					break;
1315 
1316 				/*
1317 				 * We also require super-user privilege for
1318 				 * the option.  See comments on IN6_HOPOPTS.
1319 				 */
1320 				if (!privileged)
1321 					break;
1322 
1323 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1324 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1325 				    IPPROTO_IPV6);
1326 				if (*mp)
1327 					mp = &(*mp)->m_next;
1328 				break;
1329 			case IPPROTO_ROUTING:
1330 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1331 					break;
1332 
1333 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1334 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1335 				    IPPROTO_IPV6);
1336 				if (*mp)
1337 					mp = &(*mp)->m_next;
1338 				break;
1339 			case IPPROTO_HOPOPTS:
1340 			case IPPROTO_AH: /* is it possible? */
1341 				break;
1342 
1343 			default:
1344 				/*
1345 			 	 * other cases have been filtered in the above.
1346 				 * none will visit this case.  here we supply
1347 				 * the code just in case (nxt overwritten or
1348 				 * other cases).
1349 				 */
1350 #ifdef PULLDOWN_TEST
1351 				m_freem(ext);
1352 #endif
1353 				goto loopend;
1354 
1355 			}
1356 
1357 			/* proceed with the next header. */
1358 			off += elen;
1359 			nxt = ip6e->ip6e_nxt;
1360 			ip6e = NULL;
1361 #ifdef PULLDOWN_TEST
1362 			m_freem(ext);
1363 			ext = NULL;
1364 #endif
1365 		}
1366 	  loopend:
1367 		;
1368 	}
1369 #undef IS2292
1370 }
1371 
1372 void
1373 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1374 {
1375 	struct socket *so;
1376 	struct mbuf *m_mtu;
1377 	struct ip6_mtuinfo mtuctl;
1378 
1379 	so =  in6p->inp_socket;
1380 
1381 	if (mtu == NULL)
1382 		return;
1383 
1384 #ifdef DIAGNOSTIC
1385 	if (so == NULL)		/* I believe this is impossible */
1386 		panic("ip6_notify_pmtu: socket is NULL");
1387 #endif
1388 
1389 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1390 	mtuctl.ip6m_mtu = *mtu;
1391 	mtuctl.ip6m_addr = *dst;
1392 
1393 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1394 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1395 		return;
1396 
1397 	lwkt_gettoken(&so->so_rcv.ssb_token);
1398 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1399 	    == 0) {
1400 		m_freem(m_mtu);
1401 		/* XXX: should count statistics */
1402 	} else {
1403 		sorwakeup(so);
1404 	}
1405 	lwkt_reltoken(&so->so_rcv.ssb_token);
1406 }
1407 
1408 #ifdef PULLDOWN_TEST
1409 /*
1410  * pull single extension header from mbuf chain.  returns single mbuf that
1411  * contains the result, or NULL on error.
1412  */
1413 static struct mbuf *
1414 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1415 {
1416 	struct ip6_ext ip6e;
1417 	size_t elen;
1418 	struct mbuf *n;
1419 
1420 #ifdef DIAGNOSTIC
1421 	switch (nxt) {
1422 	case IPPROTO_DSTOPTS:
1423 	case IPPROTO_ROUTING:
1424 	case IPPROTO_HOPOPTS:
1425 	case IPPROTO_AH: /* is it possible? */
1426 		break;
1427 	default:
1428 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1429 	}
1430 #endif
1431 
1432 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1433 	if (nxt == IPPROTO_AH)
1434 		elen = (ip6e.ip6e_len + 2) << 2;
1435 	else
1436 		elen = (ip6e.ip6e_len + 1) << 3;
1437 
1438 	n = m_getb(elen, M_NOWAIT, MT_DATA, 0);
1439 	if (n == NULL)
1440 		return NULL;
1441 	n->m_len = 0;
1442 
1443 	if (elen >= M_TRAILINGSPACE(n)) {
1444 		m_free(n);
1445 		return NULL;
1446 	}
1447 
1448 	m_copydata(m, off, elen, mtod(n, caddr_t));
1449 	n->m_len = elen;
1450 	return n;
1451 }
1452 #endif
1453 
1454 /*
1455  * Get pointer to the previous header followed by the header
1456  * currently processed.
1457  * XXX: This function supposes that
1458  *	M includes all headers,
1459  *	the next header field and the header length field of each header
1460  *	are valid, and
1461  *	the sum of each header length equals to OFF.
1462  * Because of these assumptions, this function must be called very
1463  * carefully. Moreover, it will not be used in the near future when
1464  * we develop `neater' mechanism to process extension headers.
1465  */
1466 char *
1467 ip6_get_prevhdr(struct mbuf *m, int off)
1468 {
1469 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1470 
1471 	if (off == sizeof(struct ip6_hdr))
1472 		return (&ip6->ip6_nxt);
1473 	else {
1474 		int len, nxt;
1475 		struct ip6_ext *ip6e = NULL;
1476 
1477 		nxt = ip6->ip6_nxt;
1478 		len = sizeof(struct ip6_hdr);
1479 		while (len < off) {
1480 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1481 
1482 			switch (nxt) {
1483 			case IPPROTO_FRAGMENT:
1484 				len += sizeof(struct ip6_frag);
1485 				break;
1486 			case IPPROTO_AH:
1487 				len += (ip6e->ip6e_len + 2) << 2;
1488 				break;
1489 			default:
1490 				len += (ip6e->ip6e_len + 1) << 3;
1491 				break;
1492 			}
1493 			nxt = ip6e->ip6e_nxt;
1494 		}
1495 		if (ip6e)
1496 			return (&ip6e->ip6e_nxt);
1497 		else
1498 			return NULL;
1499 	}
1500 }
1501 
1502 /*
1503  * get next header offset.  m will be retained.
1504  */
1505 int
1506 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1507 {
1508 	struct ip6_hdr ip6;
1509 	struct ip6_ext ip6e;
1510 	struct ip6_frag fh;
1511 
1512 	/* just in case */
1513 	if (m == NULL)
1514 		panic("ip6_nexthdr: m == NULL");
1515 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1516 		return -1;
1517 
1518 	switch (proto) {
1519 	case IPPROTO_IPV6:
1520 		if (m->m_pkthdr.len < off + sizeof(ip6))
1521 			return -1;
1522 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1523 		if (nxtp)
1524 			*nxtp = ip6.ip6_nxt;
1525 		off += sizeof(ip6);
1526 		return off;
1527 
1528 	case IPPROTO_FRAGMENT:
1529 		/*
1530 		 * terminate parsing if it is not the first fragment,
1531 		 * it does not make sense to parse through it.
1532 		 */
1533 		if (m->m_pkthdr.len < off + sizeof(fh))
1534 			return -1;
1535 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1536 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1537 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1538 			return -1;
1539 		if (nxtp)
1540 			*nxtp = fh.ip6f_nxt;
1541 		off += sizeof(struct ip6_frag);
1542 		return off;
1543 
1544 	case IPPROTO_AH:
1545 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1546 			return -1;
1547 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1548 		if (nxtp)
1549 			*nxtp = ip6e.ip6e_nxt;
1550 		off += (ip6e.ip6e_len + 2) << 2;
1551 		return off;
1552 
1553 	case IPPROTO_HOPOPTS:
1554 	case IPPROTO_ROUTING:
1555 	case IPPROTO_DSTOPTS:
1556 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1557 			return -1;
1558 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1559 		if (nxtp)
1560 			*nxtp = ip6e.ip6e_nxt;
1561 		off += (ip6e.ip6e_len + 1) << 3;
1562 		return off;
1563 
1564 	case IPPROTO_NONE:
1565 	case IPPROTO_ESP:
1566 	case IPPROTO_IPCOMP:
1567 		/* give up */
1568 		return -1;
1569 
1570 	default:
1571 		return -1;
1572 	}
1573 
1574 	return -1;
1575 }
1576 
1577 /*
1578  * get offset for the last header in the chain.  m will be kept untainted.
1579  */
1580 int
1581 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1582 {
1583 	int newoff;
1584 	int nxt;
1585 
1586 	if (!nxtp) {
1587 		nxt = -1;
1588 		nxtp = &nxt;
1589 	}
1590 	while (1) {
1591 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1592 		if (newoff < 0)
1593 			return off;
1594 		else if (newoff < off)
1595 			return -1;	/* invalid */
1596 		else if (newoff == off)
1597 			return newoff;
1598 
1599 		off = newoff;
1600 		proto = *nxtp;
1601 	}
1602 }
1603 
1604 struct ip6aux *
1605 ip6_addaux(struct mbuf *m)
1606 {
1607 	struct m_tag *mtag;
1608 
1609 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1610 	if (!mtag) {
1611 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1612 		    M_NOWAIT);
1613 		if (mtag)
1614 			m_tag_prepend(m, mtag);
1615 	}
1616 	if (mtag)
1617 		bzero(m_tag_data(mtag), sizeof(struct ip6aux));
1618 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1619 }
1620 
1621 struct ip6aux *
1622 ip6_findaux(struct mbuf *m)
1623 {
1624 	struct m_tag *mtag;
1625 
1626 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1627 	return mtag ? (struct ip6aux *)m_tag_data(mtag) : NULL;
1628 }
1629 
1630 void
1631 ip6_delaux(struct mbuf *m)
1632 {
1633 	struct m_tag *mtag;
1634 
1635 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1636 	if (mtag)
1637 		m_tag_delete(m, mtag);
1638 }
1639 
1640 /*
1641  * System control for IP6
1642  */
1643 
1644 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1645 	0,		0,		0,		0,
1646 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1647 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1648 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1649 	0,		0,		0,		0,
1650 	ENOPROTOOPT
1651 };
1652